NSFDEB-NERC: Machine learning tools to discover balancing selection in genomes from spatial and temporal autocorrelations

平衡选择是维持物种内部和物种间变异的适应性进化的关键驱动力。然而，尽管有一些值得注意的例子，人们对最近或短暂的平衡选择知之甚少，可能是因为它的基因组足迹很难与中性进化的基因组足迹区分开来。尽管人工智能的复兴已经改变了我们寻找适应性基因组区域的方式，但从当代基因组数据中描述最近平衡选择的特征仍然很困难。此外，从历史样本和非模式生物中获得的数据往往充满了技术障碍，现有方法无法解决这些障碍。因此，本提案的目标是开发一套深度学习工具，用于从时间和空间采样的基因组数据中研究近期和瞬态平衡选择，同时确保这些方法考虑到非模型研究系统中遇到的许多技术挑战。具体来说，我们将设计预测器，用于检测信号和学习从不完整、低质量和未分阶段的古代样本中进行平衡选择的参数（目标1），使用规避遗传和人口参数不确定性的方法（目标2），并扩展到由成本效益池测序策略生成的数据（目标3）。我们将把这些新工具部署到三个经验数据集，这些数据集分别包含一组最近的平衡选择案例研究，目标1、2和3解决的技术问题旨在克服这些问题。初步的研究结果支持了提出的目标的希望和可行性，我们期望这些工具为进化界提供一个强大的框架，以解决目前关于模型和非模型系统中适应的无法回答的问题。在进化生物学中，阐明物种内部变异的适应性维持的过程是一个基本问题，而现有的工具无法从非模型研究系统中存在的大量、通常是非理想的数据中解决这个问题。pi已经成功地设计了统计和机器学习方法来发现适应的足迹，并解决了关于跨几个研究系统平衡选择的目标假设。因此，通过利用非模式生物的各种数据类型特征的基因组、空间和时间自相关性，他们很好地准备发展所提出的深度学习方法来研究平衡选择。当数据不完整、低质量和未分阶段（目标1），当遗传和人口参数不确定（目标2），以及当个体水平的基因型信息不可用（目标3）时，这些方法的可用性将促进平衡选择的研究。此外，我们建议将这些方法应用于各种研究系统，以解决有关在不同时间和地理尺度上平衡选择的作用和具体模式的问题。最后，开发的工具将适用于模型生物和非模型生物中常见的广泛数据类型，通过消除数据质量限制和当前人口历史知识所造成的障碍，增强未来适应研究的能力。

OverviewBalancing selection is a key driver of adaptive evolution that maintains variation within and acrossspecies. Yet, despite a few notable examples, little is known about recent or fleeting balancing selection,likely because its genomic footprints are difficult to distinguish from those of neutral evolution. Whereasthe renaissance of artificial intelligence has transformed how we search for adaptive genomic regions, thecharacterization of recent balancing selection from contemporary genomic data remains difficult.Moreover, data obtained from historical samples and nonmodel organisms are often fraught with technicalhurdles that available methods are ill-equipped to navigate. Hence, the objective of this proposal is todevelop a suite of deep learning tools for studying recent and transient balancing selection fromtemporally and spatially sampled genomic data, while ensuring that these methods account for the manytechnical challenges encountered in nonmodel study systems. Specifically, we will design predictors fordetecting signals and learning parameters of balancing selection from incomplete, low-quality, andunphased ancient samples (Aim 1), using approaches that circumvent the uncertainty in genetic anddemographic parameters (Aim 2), and that extend to data generated by cost-efficient pooled sequencingstrategies (Aim 3). We will deploy these new tools to three empirical datasets, which respectivelyencompass a set of recent balancing selection case studies for which the technical issues tackled byAims 1, 2, and 3 are designed to overcome. Preliminary findings support the promise and feasibility of theproposed aims, and we expect these tools to provide the evolution community with a powerful frameworkto address currently unanswerable questions about adaptation in both model and nonmodel systems.Intellectual MeritElucidating the processes underlying adaptive maintenance of variation within species is a fundamentalproblem in evolutionary biology, and one for which available tools are ill-equipped to address from thevast, often non-ideal, data that exist for nonmodel study systems. The PIs have demonstrated success indesigning statistical and machine learning methods for uncovering footprints of adaptation andaddressing targeted hypotheses about balancing selection across several study systems. Thus, they arewell-poised to develop the proposed deep learning approaches for studying balancing selection, byleveraging genomic, spatial, and temporal autocorrelations across a variety of data types characteristic ofthose from nonmodel organisms. Availability of these methods will facilitate studies of balancing selectionwhen data are incomplete, low-quality, and unphased (Aim 1), when genetic and demographicparameters are uncertain (Aim 2), and when genotype information at the individual level is unavailable(Aim 3). Moreover, our proposed applications of these methods to a diversity of study systems willaddress questions regarding the roles and specific modes of balancing selection at different temporal andgeographic scales. Finally, the developed tools will be applicable to a wide range of data types commonacross model and nonmodel organisms, empowering future studies of adaptation by removing barriersimposed by limitations of data quality and current knowledge of demographic history.